Algorithmic supply of IR sensors with FPN using texture homogeneity levels

Abstract

The new algorithm for infrared (IR) image enhancement compensating fixed pattern noise (FPN) of the volt sensitivity in IR un-cooled sensors is described. The main idea of the algorithm lies in the presentation of any input image as the set of calibrating segments. The segmentation procedure uses the texture homogeneity levels and the estimators of the transfer coefficients calculated within each level. The results of modeling have shown the advantage of the proposed algorithm comparing with well-known 2-point FPN correction procedures under PSNR and SSIM criteria.